NVAI-open
Make images, videos with sound, voiceovers and 3D models on your own computer — no subscription, no cloud uploads.
NVAI-open is a local AI media tool that runs on your own computer: images, videos with sound, music, voiceovers and 3D models all come out of your own graphics card, with no account, no monthly subscription and nothing sent to the cloud. The author built it for artwork for their own game projects, the repository has 17 stars, and version 0.5.21 for Windows shipped in November 2026. It covers several jobs: images from local models such as FLUX and Qwen-Image, with side-by-side comparisons of models on the same picture and seed; videos with sound, including turning a photo into a short talking clip; voiceovers from Qwen3-TTS in 10 languages, where one sentence sets a speaker's voice; and 3D meshes straight from an image. A Studio view stitches picture, sound and music into one timeline for export. Setup is install the app, then download models inside it — after that everything runs locally. The Windows build wants an NVIDIA or AMD graphics card and the Mac build wants an Apple Silicon Mac; the author tested on a Windows machine with an RTX 5060 Ti and a Mac mini M1. There is also an interface for Claude Code, so an AI assistant can ask it for assets while it works. Three things to know: you download the models yourself, so disk space and graphics power matter; the author maintains it in their spare time, so replies can be slow; and one video model, MiniMax H3, carries a licence that forbids use in the EU, the UK, South Korea and the USA without a separate licence from MiniMax, and forbids publishing anything made with it — the app marks it in red, so read that warning first.
@Dominic F.